Skip navigation
Please use this identifier to cite or link to this item:
Title: Efficient plant leaf representations: A comparative study
Authors: Prasad S.
Peddoju, Sateesh Kumar
Ghosh D.
Published in: Proceedings of IEEE Region 10 Annual International Conference
Abstract: This paper studies various different ways of representing plant leaves in different feature space to recognize them accurately and efficiently in changing scenario and dimension. Here, leaves are represented as: (i) statistical features - relative sub-image sparse coefficient (RSSC), (ii) shape features - angle view projection (AVP), (iii) multi-resolution features - Gabor Wavelet transform combined with gray level co-occurrence matrix (GWT-GLCM), (iv) deep features - convolutional neural network (CNN) and (v) hand-crafted ResNet features. It is experimentally found that all the above proposed algorithms outperform the state-of-the-arts and the hand-crafted Gabor-ResNet feature space among them is the superior approach. This study also help researchers in analyzing the pattern distributions of the same object in different feature spaces. © 2017 IEEE.
Citation: Proceedings of IEEE Region 10 Annual International Conference, (2017), 1175- 1180
Issue Date: 2017
Publisher: Institute of Electrical and Electronics Engineers Inc.
Keywords: Neural networks
Wavelet transforms
Comparative studies
Convolutional Neural Networks (CNN)
Efficient plants
Gabor wavelet transforms
Gray level co-occurrence matrix
Multi-resolution feature
State of the art
Statistical features
Plants (botany)
ISBN: 9.78151E+12
ISSN: 21593442
Author Scopus IDs: 36997272500
Author Affiliations: Prasad, S., Biometrics and Forensics Lab, NTU, Singapore
Peddoju, S.K., HPC Lab, Department of CSE, IIT Roorkee, India
Ghosh, D., SP Lab, Department of ECE, IIT Roorkee, India
Appears in Collections:Conference Publications [CS]

Files in This Item:
There are no files associated with this item.
Show full item record

Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.